Artificial intelligence is enlarging the wealth pool while also making income distribution more uneven, according to the Quarterly Report on Macroeconomic Policy (Q2 2026) released by the China Finance 40 Forum, or CF40, on July 27.
The report says U.S. household directly held equity wealth climbed from about $29 trillion in 2019 to roughly $55 trillion in the first quarter of 2026, nearly doubling over the period. It adds that the fastest expansion came after 2023, when AI-driven gains in stock prices took hold.
CF40 estimates that from 2022 to the first quarter of 2026, directly held equity wealth among U.S. households increased by about $21 trillion. Of that amount, the top 10% of households captured around 88%, or $18.5 trillion, while the bottom 50% received about 1%, or $0.2 trillion. The report says the shift offers an early picture of how AI-related technological gains can coincide with a more imbalanced distribution of income and wealth.
Huang Yiping links AI to earlier episodes of distributional change
Speaking at the release event, Huang Yiping, a CF40 member and dean of the National School of Development at Peking University, said every major technological advance in history has had a deep effect on the structure of income distribution.
He cited research by economic historian Robert Allen on the so-called “Engels’ Pause” during the first Industrial Revolution. In that period, output per worker rose for decades after the industrial breakthrough, but workers’ real wages did not show a clear increase. In terms of distribution, labor income did not begin to recover as a share of total output until about 100 years after the start of the first Industrial Revolution.
That history, Huang said, raises a question for the present AI cycle: whether the early phase of AI progress could bring a similar stretch marked by stagnant wages, stronger capital returns and wider inequality.
He described AI as a general-purpose technology and a historic opportunity for China’s economic development. At the same time, he said, AI-driven technological progress could push the income distribution structure further out of balance for a period ahead. “We should attach great importance to the issue of income distribution and prepare in advance,” Huang said. He added that distribution matters for aggregate demand, and if current trends continue, the pattern of strong supply and weak demand may not be reversed in the short term and could become more pronounced, affecting the sustainability of growth.
On that basis, Huang called for a people-centered policy framework built around investment in people, with the aim of balancing short-term social stability and long-term economic efficiency so that the gains from technology are shared more broadly.
Employment and income figures show an uneven shift
CF40 says AI is already changing the balance between capital spending and labor hiring. U.S. market data cited in the report show that from 2024 to 2025, employment growth in sectors with relatively high AI adoption, including information industries, professional services, and finance and insurance, was markedly below the average pace recorded from 2010 to 2019.
At the same time, AI contributed to a sharp increase in U.S. stock market value and highly uneven gains in household wealth. The report again points to the same split: about 88% of the wealth increase went to the top 10% of households, while about 1% went to the bottom 50%.
Disposable income also showed signs of divergence. In 2025, the top 10% of households by income accounted for 36% of disposable income, above the 2010-2019 average of 34.3%. The bottom 50% of lower- and middle-income households accounted for 11.5%, slightly below the 2010-2019 average of 11.8%.
Huang said large swings in distribution following technological progress are not new. He again referred to Robert Allen’s work, which shows that the first Industrial Revolution began in the 1760s, but the share of labor wages in economic output kept falling until it stabilized and turned upward in the 1870s. In other words, it took about a century after productivity surged for labor’s share of the gains to start expanding.
He said the first Industrial Revolution produced the inequality-widening phenomenon described as “Engels’ Pause,” with capital owners taking a larger share of wealth. The second Industrial Revolution favored investment in intangible assets such as technical knowledge and organizational capital. The third Industrial Revolution brought job polarization and wage inequality: high-paying, high-skill positions expanded quickly, lower-end jobs remained relatively stable, and middle-tier jobs shrank while wage growth for medium-skill workers stalled.
Four channels through which AI may reshape distribution
Huang said AI, as a widely recognized general-purpose technology, is likely to trigger another leap in human productivity. He outlined four channels through which that process may affect income distribution.
Capital bias
The first is a capital-biased mechanism, reflected in a continued decline in labor’s share of income. Companies may keep increasing capital investment to replace traditional labor inputs, changing both factor allocation and the distribution of returns. In Huang’s formulation, AI can raise output per worker without guaranteeing that workers’ incomes rise along with it.
Task allocation and K-shaped divergence
The second is a task-based mechanism, with effects resembling the hollowing out of the middle class and the K-shaped split associated with the third Industrial Revolution. “After AI technology is implemented, are you replaced or empowered?” Huang said. “If AI empowers you, you will have more opportunities in the future and your income may move upward along the K-shape. If you are easily replaced by AI technology, your future income or returns may move downward.”
Skill polarization and the digital divide
The third is skill differentiation and the digital divide, which may block social mobility. Differences in technological capability can be translated directly into income gaps and unequal development opportunities, while also deepening industry and regional divergence and creating winner-take-all effects. “Put simply, either you have a platform or you have skills. Otherwise, the development of new technology may not be particularly favorable to you,” Huang said.
Wealth amplification
The fourth is a wealth amplification mechanism, mainly visible in the decoupling of capital returns from labor returns. Through asset appreciation, cost pass-through and intergenerational transfer of resources, inequality in wealth can widen from the capital side and produce a pattern in which the rich get richer. Huang added that this is “a general phenomenon and not strongly tied to AI.”
Why stronger productivity may not guarantee faster growth
The report then turns to a broader macro question: does a rise in productivity necessarily lead to faster economic growth?
Zhang Bin, the lead author of the report, a senior researcher at CF40 and deputy director of the Institute of World Economics and Politics at the Chinese Academy of Social Sciences, said the answer may be no. He said AI could leave demand growing more slowly than before, while overall growth is often determined by whichever side, supply or demand, is weaker.
The report quotes an academic paper published in 2020 saying that workers rely mainly on wage income and have a significantly higher marginal propensity to consume than capital owners who depend on asset appreciation. If AI shifts wealth from labor to capital owners, society could face a serious shortfall in aggregate consumer demand.
It also notes that China’s economy has shown a persistent pattern of strong supply and weak demand in recent years. In the first half of 2026, China’s real GDP growth was 4.7%. Retail sales of consumer goods rose 1.3% year over year, while nationwide fixed-asset investment excluding rural households fell 5.7%, indicating that domestic demand remained weak overall.
Huang said the supply-side boost from AI is already clear, but if aggregate demand does not recover, excessive supply strength would still leave growth unsustainable. “If the current trend continues, a scenario may emerge in which strong supply and weak demand not only cannot be reversed in the short term, but may become even more acute,” he said. “We need to think about how to keep aggregate demand rising so that it remains relatively balanced with supply.”
Three policy responses: defense, empowerment and rebalancing
Huang said China should treat AI as a historic opportunity while preparing for the possibility that income distribution becomes more distorted in the period ahead. He proposed a three-part response.
Defense
The first is defense: strengthening the social safety net to cushion AI-related job losses while curbing excess monopoly gains generated through algorithms, in order to preserve a basic sense of fairness.
- Improve unemployment insurance and minimum protection systems.
- Strengthen antitrust enforcement in the platform economy and AI-related sectors.
- Set up ethical review and restriction mechanisms for AI applications that are purely substitutional.
Huang said it is relatively certain in the short term that AI innovation will hit some occupations hard and cause some workers to lose their jobs, just as earlier industrial revolutions did. “This is a normal phenomenon, but the key issue is how to achieve a smooth transition,” he said. He argued for an employment-first principle, more support for AI innovation that empowers rather than replaces labor, and stronger basic social protection to avoid major social disruptions.
Empowerment
The second is empowerment: reshaping human capital so that workers move from being replaced by AI to being able to direct and work with AI-driven innovation.
- Reform education systems to include AI literacy and creative thinking.
- Build a lifelong vocational training system.
- Promote “AI plus occupation” certification and employment support.
Huang said formal school credentials may become less important over time for young people seeking jobs and career advancement, while hybrid skills may become more important. He said a key part of the government’s “investing in people” approach is to cultivate the ability to work collaboratively with AI.
Rebalancing
The third is rebalancing: redesigning factor distribution so that capital and technology do not monopolize AI gains, and so that the value created by data and algorithms can reach a broader part of society through institutional arrangements.
- Explore adjustment taxes on excess AI profits.
- Clarify data property rights and promote the social sharing of public data.
- Create an AI dividend distribution fund for broad public sharing.
“How to let the whole society share the value created by technology and algorithms places a very high demand on public policy,” Huang said. He added that it may still be too early to establish a universal AI dividend fund now, though support for lower-income groups in the form of living or income assistance could be considered.
Huang said inequality in overall income distribution is already a prominent problem and may become more pronounced, while insufficient demand is the most prominent issue at present. Turning “investment in people” into a full policy package, he said, could help ease the current imbalance of strong supply and weak demand and improve income distribution.
The original article was written by a Caijing reporter. It was published via the WeChat account “Caijing May Flower” (ID: Caijing-MayFlower), with Tang Jun as author and Zhang Wei as editor.